What is the diffrence bewtween NLU based Vitual agent and LLM Based Virtual agent?

[jay97]

Hello Guys, 

I wanted to know the difference between NLU based and LLM Based Virtual agent ? 

I could see in the yokohama relase boths are available ? 

Both work independently or they working togheter  ?

Please guice me on it .

Regards,

Jay

[pramodhm112]

Hello @jay97 

Here's the difference between NLU and LLM Virtual Agents:

NLU-based is the traditional approach - you have to set up intents, utterances, and entities and manage everything in the NLU Workbench. It's more hands-on and requires a lot of configuration.

LLM-based is the newer AI-powered way - you just write descriptions in normal language, and the AI handles understanding what users want. Way less work to maintain!

While both are available in the Yokohama release, they cannot work together as they're completely separate systems. You can't mix NLU topics with LLM topics; they each run their own chat client. If you turn on the LLM Virtual Agent on a portal, it replaces the NLU one. Now Assist in Virtual Agent - What's New, what's different, and what's coming! (June 2024)

https://www.servicenow.com/docs/bundle/yokohama-conversational-interfaces/page/administer/virtual-ag...

Moving from NLU to LLM: ServiceNow provides a way to convert your old NLU topics to LLM topics automatically. It's been available since Washington Patch 3. Now Assist in Virtual Agent: Migrating NLU topics to LLM

Most companies are moving to LLM because it's easier to manage and gives more natural conversations. You'll need to pick one or the other for each portal though - can't have both running at the same time.

Please let me know if you have any further questions

Thanks,
Pramodh

[kaushal_snow]

@jay97 ,

In the Yokohama release, ServiceNow introduced two distinct Virtual Agent models: the traditional NLU based Virtual Agent and the newer LLM based Virtual Agent....The NLU based Virtual Agent utilizes predefined intents and entities to interpret user inputs, relying on structured training and configuration. Conversely, the LLM based Virtual Agent, powered by large language models, offers enhanced flexibility by understanding and generating responses based on context, without the need for extensive configuration........Both models can operate independently, but they are not designed to work simultaneously within the same chat session. The NLU-based Virtual Agent uses keyword-based topics, while the LLM based Virtual Agent employs LLM topics. These two types of topics are not compatible with each other. Therefore, organizations must choose between the two based on their specific requirements.....

To integrate the LLM based Virtual Agent, the Now Assist in Virtual Agent plugin must be activated, and the assistant should be connected to the desired portal and Virtual Agent. This setup enables the Virtual Agent to leverage generative AI capabilities, enhancing user interactions with more dynamic and context aware responses.....

If you found my response helpful, please mark it as ‘Accept as Solution’ and ‘Helpful’. This helps other community members find the right answer more easily and supports the community.